[PATCH] serial/efm32: add new driver

[Alan Cox]

On Wed, 21 Dec 2011 16:05:59 +0100
Uwe Kleine-König <u.kleine-koenig at pengutronix.de> wrote:

> Signed-off-by: Uwe Kleine-König <u.kleine-koenig at pengutronix.de>
> ---
> Hello,
> 
> note that ARCH_EFM32 isn't in mainline yet, so to be actually usable
> some arch patches are needed.

Start by running it through the CodingStyle scripts as I see a //
comment in there 8)


> +static void efm32_usart_rx_chars(struct efm32_usart_port *efm_port)
> +{
> +	struct uart_port *port = &efm_port->port;
> +	struct tty_struct *tty = port->state->port.tty;

Needs to be using krefs and checking the tty is not NULL
(tty_port_tty_get)

> +static void efm32_usart_set_termios(struct uart_port *port,
> +		struct ktermios *new, struct ktermios *old)
> +{
> +	struct efm32_usart_port *efm_port = to_efm_port(port);
> +	unsigned long flags;
> +	unsigned baud;
> +	u32 clkdiv;
> +
> +	/* no modem control lines */
> +	new->c_cflag &= ~(HUPCL | CRTSCTS | CMSPAR);

Minor item - HUPCL shouldn't get cleared - its a request for hangup
behaviour not a port feature.

> +	/* currently only some features are implemented */
> +	new->c_cflag &= ~CSIZE;
> +	new->c_cflag |= CS8;
> +	new->c_cflag |= CSTOPB;
> +	new->c_cflag &= ~PARENB;

If you can do CS8 without parity you can do CS7 with parity.

> +	new->c_iflag = 0;

This seems broken. Lots of the iflgs are user requests and stack
properties not port features. For example you can do xon/xoff as its
pure software. As far as I can see you should leave c_iflag alone.

Alan